The basic idea of this paper comes from the paper published in American Political Science Review in 2022, Education or Indoctrination? The Violent Origins of Public School Systems in an Era of State-Building. Paglayan (2022) argues that states, in particular non-democracies, experienced mass violence, such as civil war, are more likely to install educational system for the children to enhance their compliance to the regime. Thus, the study examines the relationship between regime type, civil war experience, and education expansion in Latin America.
However, there are several points in Paglayan (2022) that are unresolved or unclear. First, do authoritarian regimes use indoctrination to protect children from potential mass violations following civil war? So, what are the criteria for determining the target of indoctrination? How do autocracies use indoctrination strategically to achieve their goals? Second, is school enrollment a good indicator of indoctrination?
Although the article contains insightful arguments, my question focuses on education and indoctrination. Is it true that they are mutually exclusive? Autocrats can use education to increase compliance by extending it to their children. However, it can also provide an opportunity for children to grow in their own abilities. Thus, I believe that what distinguishes education from indoctrination is determined on the demand side rather than the supply side. If the system’s beneficiaries (or indoctrinated) children have developed their capability, driving the country forward, even indoctrination strategy works as an education.
Theoretical Expectation
Education and indoctrination, I contend, are not mutually exclusive. The outcomes of the strategy are what justify education or indoctrination. For example, if the goal of an indoctrination strategy is to improve students compliance, it is less likely to provide diverse content and curriculum to improve students’ creativity. Otherwise, if such benefits are provided, even indoctrination will encourage students to have highly professionalized and developed skills, and technologies, which can have an productive impact on the entire regime.
Therefore, here, I expect the relationship between level of indoctrination, education expenditure, economic development, and intellectual property of the state, which cannot be obtained without well-designed education.
Preliminary Exploration
Intellectual property
To measure the intellectual property, I use the country-year patent data from 1995 to 2023 from World Intellectual Property Organization. What I use is the patent publication data, which are measured as total count by applicant’s origin by technology.
I transformed the data into country-year structure, manipulating COW country codes. Then, I removed the observations that do not have COW country codes from the sample.
In the sample, WIPO provides Field of technology that the published patents should belong to. There are 35 different technologies of patents in the sample. I rely on Concept of a Technology Classification for Country Comparisons, a report published in WIPO.
# ICT / Artificial Intelligence / Biotechnology / Nanotechnology / Environment-related technologies / Health
## I used a combination of different criteria to classify the technologies into low-tech,
## middle-level techs, and high-techs. These criteria include: Complexity and sophistication of the
## technology: Generally, the more complex and sophisticated the technology is, the higher the level of tech it belongs to.
## Level of innovation and research and development (R&D) involved: High-techs are typically
## associated with cutting-edge innovation and intensive R&D activities, while low-techs are
## often considered more mature and established technologies.
## Economic impact and industrial relevance: High-techs are often associated with strategic industries
## and high-value-added products, while low-techs are often associated with more traditional and
## low-value-added sectors.
## Based on these criteria, I classified the technologies into three broad categories as follows:
## Low-tech: These are technologies that are relatively simple, mature, and established, and
## typically involve low levels of innovation and R&D. They are often associated with traditional
## and low-value-added sectors such as construction, transportation, and textiles.
## Middle-level techs: These are technologies that are more complex and sophisticated than low-techs
## but not as cutting-edge and innovative as high-techs. They typically involve
## moderate levels of innovation and R&D and are associated with a wide range of
## sectors, such as manufacturing, healthcare, and logistics.
## High-techs: These are technologies that are characterized by high levels of innovation and R&D,
## cutting-edge research and development, and strategic importance for the economy.
## They are typically associated with high-value-added industries such as biotechnology,
## information technology, and telecommunications.
patent_long |> mutate(
Class = case_when(
`Field of technology` %in% c("Basic communication processes", "Basic materials chemistry",
"Civil engineering", "Furniture, games", "Handling", "Textile and paper machines",
"Thermal processes and apparatus", "Transport") ~ "Low-tech",
`Field of technology` %in% c("Audio-visual technology", "Chemical engineering", "Electricity",
"Engines, pumps, turbines", "Food chemistry", "Machine tools",
"Measurement", "Mechanical elements", "Medical technology",
"Optical instruments", "Packaging", "Paper and printing",
"Surface technology, coating") ~ "Middle-tech",
`Field of technology` %in% c("Analysis of biological materials", "Biotechnology", "Computer technology",
"Control", "Digital communication", "Electrical machinery, apparatus, energy",
"Environmental technology", "IT methods for management",
"Macromolecular chemistry, polymers", "Materials, metallurgy",
"Micro-structural and nano-technology", "Optics", "Organic fine chemistry",
"Other consumer goods", "Other special machines", "Pharmaceuticals",
"Semiconductors", "Telecommunications") ~ "High-tech",
T ~ NA_character_
)
) -> patent_long
patent_long |> group_by(ccode, year) |> dplyr::filter(!is.na(no_patent)) |>
count(`Field of technology`) |>
summarize(diversity = sum(n, na.rm = T)) -> patent_diversity
patent_long |>
group_by(Origin, ccode, year) |>
summarize(
sum_patent = sum(no_patent, na.rm = T)) -> patent_long_sum
patent_diversity_long <- patent_long_sum |> left_join(
patent_diversity, by = c("ccode", "year")
)
patent_diversity_long <- patent_diversity_long |> mutate(diversity = replace_na(diversity, 0))
patent_long |> left_join(
patent_diversity_long, by = c("ccode", "year")
) -> sample
vindoc_cy_full |>
dplyr::select(country_name, year, v2xed_ed_inpt, v2xed_ed_poed, v2xed_ed_inco, v2xed_ed_cent,
v2edmath,
e_gdppc, e_gdp, resource_rev = e_total_resources_income_pc) |>
mutate(ccode = countrycode::countrycode(country_name, "country.name", "cown")) ->
vindoc_sample
vdemdata::vdem -> vdem
vdem |> dplyr::select(COWcode, year, v2x_polyarchy, v2x_libdem,
e_boix_regime) -> vdem_sample
sample |>
mutate(year = as.numeric(year)) |>
left_join(vindoc_sample,
by = c("ccode", "year")) |>
left_join(vdem_sample,
by = c("ccode" = "COWcode", "year")) ->
patent_indoc
patent_indoc |> group_by(ccode) |>
mutate(lag_patent = dplyr::lag(sum_patent, n = 1, order_by = year),
lag_gdp = dplyr::lag(e_gdp, n = 1, order_by = year),
diff_gdp = e_gdp - lag_gdp,
gdpgrth = diff_gdp/lag_gdp) -> patent_indoc
table(patent_indoc$Class)
lm(log(sum_patent+1) ~ v2xed_ed_inpt + I(v2xed_ed_inpt^2) + gdpgrth + log(lag_patent + 1) +
resource_rev + e_gdppc + v2edmath + e_boix_regime +
#v2x_libdem +
as.factor(ccode) + as.factor(year),
data = patent_indoc |>
dplyr::filter(Class %in% "Low-tech")) |>
# texreg::screenreg(omit.coef = "as.factor")
ggeffects::ggpredict(terms = c("v2xed_ed_inpt")) |> plot()
lm(log(sum_patent+1) ~ v2xed_ed_inpt + I(v2xed_ed_inpt^2) + gdpgrth + log(lag_patent + 1) +
resource_rev + e_gdppc + v2edmath + e_boix_regime +
#v2x_libdem +
as.factor(ccode) + as.factor(year),
data = patent_indoc |>
dplyr::filter(Class %in% "Middle-tech")) |>
# texreg::screenreg(omit.coef = "as.factor")
ggeffects::ggpredict(terms = c("v2xed_ed_inpt")) |> plot()
lm(log(sum_patent+1) ~ v2xed_ed_inpt + I(v2xed_ed_inpt^2) + gdpgrth + log(lag_patent + 1) +
resource_rev + e_gdppc + v2edmath + e_boix_regime +
#v2x_libdem +
as.factor(ccode) + as.factor(year), data = patent_indoc |> dplyr::filter(Class %in% "High-tech")) |>
#texreg::screenreg(omit.coef = "as.factor")
ggeffects::ggpredict(terms = c("v2xed_ed_inpt")) |> plot()
lm(diversity ~ v2xed_ed_inpt + I(v2xed_ed_inpt^2)+ gdpgrth + log(lag_patent + 1) +
resource_rev + e_gdppc + v2edmath + e_boix_regime +
#v2x_libdem +
as.factor(ccode) + as.factor(year),
data = patent_indoc) |>
# texreg::screenreg(omit.coef = "as.factor")
ggeffects::ggpredict(terms = c("v2xed_ed_inpt")) |> plot()
library(janitor)
patent_indoc |> ungroup() |>
mutate(
indoc = if_else(v2xed_ed_inpt <= quantile(v2xed_ed_inpt, probs = 0.2, na.rm = T), "Low_indoc",
if_else(v2xed_ed_inpt >= quantile(v2xed_ed_inpt, probs = 0.8, na.rm = T), "High_indoc", "Middle_indoc")),
indoc = factor(indoc, levels = c("Low_indoc", "Middle_indoc", "High_indoc"))) |>
group_by(Class, indoc) |> drop_na(indoc) |>
summarize(mean_patent = mean(sum_patent, na.rm = T))